Euphemism is a pervasive linguistic phenomenon through which speakers replace direct, potentially offensive, sensitive, or socially undesirable expressions with alternative forms that reduce their negative pragmatic impact. Although euphemism has received considerable attention in linguistic, sociolinguistic, pragmatic, and discourse-analytic research, comparatively less attention has been paid to the ways in which euphemistic expressions are formally constructed across typologically diverse languages. This study investigates cross-linguistic patterns of euphemism formation in political discourse through a corpus-based comparative framework. The proposed multilingual corpus focuses on English, Uzbek, and Russian political discourse and examines euphemistic expressions at lexical, semantic, metaphorical, metonymic, morphological, and grammatical levels. The study adopts an integrated theoretical framework combining euphemism theory, pragmatics, political discourse analysis, and corpus linguistics. Euphemistic expressions are classified according to the linguistic mechanism through which a direct nomination is transformed into a socially or politically acceptable alternative. Particular attention is given to metaphor, metonymy, generalization, understatement, periphrasis, abstraction, lexical substitution, grammatical restructuring, and abbreviation. The analysis demonstrates that euphemization is simultaneously constrained by universal pragmatic motivations and shaped by language-specific structural resources and sociocultural conventions. Metaphorical and periphrastic formation emerge as particularly productive mechanisms in political discourse, while the distribution of morphological and grammatical strategies varies according to the structural characteristics of individual languages. The findings suggest that political euphemism should not be treated merely as lexical substitution but as a multidimensional process involving linguistic form, conceptual representation, pragmatic intention, and ideological positioning. The study contributes to cross-linguistic euphemism research by proposing a corpus-based classification framework that can support systematic comparison across languages.